Eolian cannibalism: Reworked loess and fluvial sediment as the … · 2018. 8. 7. · Reworked loess and fluvial sediment as the main sources of the Chinese Loess Plateau Geological

Licht et al.

944 Geological Society of America Bulletin, v. 128, no. 5/6

Eolian cannibalism: Reworked loess and fluvial sediment as the main sources of the Chinese Loess Plateau

A. Licht1,2,3,†, A. Pullen1,4, P. Kapp1, J. Abell1, and N. Giesler1

1Department of Geosciences, University of Arizona, Tucson, Arizona 85721, USA2Biodiversity Institute, University of Kansas, Lawrence, Kansas 66045, USA3Institut für Erd- und Umweltwissenschaften, Potsdam Universität, 14476 Potsdam, Germany4Department of Earth and Environmental Sciences, University of Rochester, Rochester, New York 14627, USA

ABSTRACT

The loess and paleosol sequences of the Chinese Loess Plateau are composed of Qua-ternary dust, the origin of which has been the subject of considerable debate. Some recent U-Pb geochronological studies of eo-lian zircons have proposed the existence of two major wind pathways: from the north and northwest, through the Badan Jaran, Tengger, and Mu Us Deserts during inter-glacials, and from the west, through the Qaidam Basin during glacials. Others have emphasized the importance of Yellow River sediment supply in the Chinese Loess Plateau sediment budget. However, tracking dust source regions through U-Pb dating in a sta-tistically robust manner is particularly com-plex given the similar age peaks in the age probability distributions of potential source regions in Central Asia. This paper pre sents 2410 new U-Pb ages of detrital zircons from wind-eroded strata, Quaternary eolian de-posits, and modern river sands in central China in order to increase the robustness and the spatial resolution of zircon age distribu-tions in dust source regions. We then propose a new mixture modeling technique to statisti-cally address the contribution of these differ-ent sources to the Chinese Loess Plateau sedi-mentary budget. Our contribution estimates indicate that eolian supply to the Chinese Loess Plateau is dominated (60%–70%) by reworking of Yellow River sediment. More-over, evidence of Qaidam Basin–sourced zir-cons (15%–20%) in both loess (glacial) and paleosol (interglacial) layers corroborates the existence of an erosive wind pathway through the Qaidam Basin during glacials and implies that a substantial portion of the interglacial dust is recycled from older glacial loess. We

propose that sediment reworking of Yellow River sediment and older loess deposits by wind on the Chinese Loess Plateau homoge-nized the eolian zircon populations toward a glacial provenance due to higher (2–20 times) dust accumulation rates during gla-cials. These findings suggest that the Chinese Loess Plateau has evolved as a more dynamic landform than previous thought, where wind deflation, fluvial input, lateral transport, and accumulation of sediment are equally impor-tant. These internal reworking effects would then significantly bias the paleoclimatic inter-pretations based on eolian dust properties of the Chinese Loess Plateau.

INTRODUCTION

The Chinese Loess Plateau (Fig. 1) is mainly composed of eolian dust deposits that are pre-dominantly brought by spring storms sweeping clastic material from the deserts of the Asian interior (Roe, 2009). Though the oldest loess deposits within the southernmost and western-most Chinese Loess Plateau have been dated to 8 Ma (An et al., 2001), 22–25 Ma (Guo et al., 2002; Qiang et al., 2011), and 41 Ma (Licht et al., 2014), most of the Chinese Loess Pla-teau is composed of Quaternary loess-paleosol sequences. Loess layers were deposited during glacial periods under a colder and drier Asian climate, whereas paleosol layers developed dur-ing warmer and moister interglacial periods with enhanced summer monsoons (An et al., 1990; Porter and An, 1995; Kang et al., 2011, 2013).

Loess deposits consist of a coarse (12–70 mm) primary fraction, transported via saltation by dust storms in near-surface suspension clouds, and a fine (1–3 mm) secondary fraction, the ori-gin of which is still controversial, either pedo-genic, supplied by upper-level airflow, and/or by adherence to the coarse fraction (Derbyshire et al., 1998; Sun et al., 2006; Stevens and Lu,

2009; Vandenberghe, 2013). In the modern inter-glacial climate, dust storms originate north and northwest of the Chinese Loess Plateau, in the Gobi Desert, commonly during the breakdown of the Siberian High in Spring (Roe, 2009), and they track across the Badan Jaran, Tengger, and Mu Us Deserts (Wang et al., 2004). These des-ert areas are thought to be the main source areas for the coarse fraction of the modern dust to the plateau, though the exact locus of dust genera-tion in these regions—piedmont alluvial fans, dried lake basins, or sand dune fields—is widely debated (Pye, 1995; Sun, 2002; Amit et al., 2014). Additionally, recent provenance studies of loess have also highlighted the importance of reworking of local substratum rocks (mostly pre-Quaternary terrestrial sediment) and of flu-vial detritus brought by the Yellow River, with headwaters in northeastern Tibet (Fig. 1), in the loess sedimentary budget (Stevens et al., 2013; Che and Li, 2013; Bird et al., 2015; Nie et al., 2015; Kapp et al., 2015).

The geographic origin of pre-Holocene gla-cial loess is more controversial. Wind-eroded landforms in the Qaidam Basin (Kapp et al., 2011; Rohrmann et al., 2013), U-Pb geochro-nology of eolian zircons (Pullen et al., 2011), and grain-size data and climate simulations (Vandenberghe et al., 2006) have suggested the existence of a different dust-storm track during glacial periods. These observations point to the loci of source areas located south of the Qilian Shan mountain ranges, in the Qaidam Basin and northern Tibetan Plateau (Fig. 1). This arid region would have been deflated by surface westerly winds, in response to either enhanced penetration of Atlantic westerlies during glacial periods (Vandenberghe et al., 2006), and/or to the shift of the whole midlatitude atmospheric system by ~10° of latitude equatorward (Togg-weiler and Russell, 2008; Kapp et al., 2011). In deeper time, this westerly wind pathway would have been favored by a lower-elevation north-

GSA Bulletin; May/June 2016; v. 128; no. 5/6; p. 944–956; doi: 10.1130/B31375.1; 9 figures; 2 tables; Data Repository item 2016026; published online 8 January 2016.

†licht@ uw .edu

For permission to copy, contact [email protected] © 2016 Geological Society of America

mailto:alicht%40ku.edu?subject=

Reworked loess and fluvial sediment as the main sources of the Chinese Loess Plateau

Geological Society of America Bulletin, v. 128, no. 5/6 945

ern Tibetan Plateau until the late Neogene–early Quaternary (Sun et al., 2008; Nie et al., 2014). However, the relative contributions from the Qaidam Basin compared with that of the more proximal desert regions and the Yellow River to the north in the eolian sedimentary budget are poorly constrained, given the complexity of the provenance problem with multiple potential source regions.

Single-grain U-Pb dating of detrital zircons is an efficient and effective technique to dif-ferentiate the contribution of discrete sources in sediment samples (Gehrels, 2000). U-Pb analy ses of detrital zircon grains produce age distributions that reflect the ages of zircons in the source rocks for the sediment. The pres-ence or absence of a particular age group in a sample is used as a test to determine if a par-ticular source region contributes to the sample (Gehrels et al., 2011). Recent developments in laser-ablation techniques allow for U-Pb analy-ses of small (<40 µm) zircons in eolian dust and application of this method to Chinese loess deposits (Stevens et al., 2010, 2013; Pullen et al., 2011; Xiao et al., 2012; Che and Li, 2013; Nie et al., 2014; Bird et al., 2015). Tracking Chinese loess dust source regions through U-Pb dating is particularly complex given the similar age peaks in the age probability distributions of western and northern deserts in central China, as well as in individual mountain ranges and ter-ranes ( Gehrels et al., 2003, 2011; Stevens et al., 2010). Central Asian age distributions of detrital and igneous rocks are commonly dominated by two major age groups, 225–330 Ma and 380–500 Ma, with two secondary, broader age popu-lations, 750–1000 Ma and 1500–2800 Ma, the relative contribution of which varies regionally (Gehrels et al., 2003; Stevens et al., 2010; Pullen

et al., 2011; Lease et al., 2007, 2012; Weislogel et al., 2010). However, the number (n) of U-Pb ages per published loess sample (commonly n < 150) is so far too low to accurately constrain the relative contribution of each age group (see next section; e.g., Andersen, 2005; Pullen et al., 2014). Some workers have been using the multi-dimensional scaling (MDS) statistical technique of Vermeesch (2013), based on the less n-depen-dent Kolmogorov-Smirnov (KS) statistic, as a visualization tool to qualitatively estimate the similarity (or dissimilarity) between loess sam-ples and potential sources (Stevens et al., 2013; Che and Li, 2013; Bird et al., 2015). However, this approach does not provide quantitative esti-mates of the relative contribution of the different dust source regions.

This paper presents new U-Pb ages of detrital zircons from eolian and fluvial deposits as well as from wind-eroded strata in central China, in order to better illuminate the zircon age dis-tributions in potential dust source regions. We then present a new mixture modeling technique that statistically determines the contribution of different sources based on U-Pb data, and we use it to assess the provenance of the Chinese Loess Plateau.

SAMPLES AND METHODS

Sampling Strategy

We first expanded the data set of existing U-Pb ages surrounding the Chinese Loess Pla-teau with 23 new samples to capture regional variability in eolian sediment source. Surface samples of up to 4 kg were collected from vari-ous locations in central China (Fig. 2). In the arid areas north and west of the Chinese Loess

Plateau, we sampled modern eolian sand dunes (one sample in the Mu Us Desert, four in the Tengger Desert, and two in the Badan Jaran Desert). We also sampled the tops of yardangs (three samples in the Mu Us Desert and four in the Badan Jaran Desert), streamlined landforms sculpted by windblown sand and composed of Pliocene–Quaternary to Cretaceous sedimen-tary bedrock. In the western, deflationary, and bedrock-floored Qaidam Basin, we sampled Miocene–Pliocene strata that have been sculpted into yardangs (three samples, which were previ-ously assessed using cosmogenic nuclides by Rohrmann et al., 2013). To represent the aver-age influx of sediment from northeast Tibet, we sampled modern fluvial deposits from the Yellow River at the western edge of the Mu Us Desert (one sample). We also determined new U-Pb ages on zircons from paleosol layers in the Heimugou section (layers S0, S1, S9, S15 and S22; after Porter, 2001) of the central Chinese Loess Plateau.

We determined U-Pb ages on ~70 to ~370 detrital zircons per sample. In total, we increased the existing data set of Central Asian U-Pb ages by 2410 ages from 23 samples. The nature, location, and number of U-Pb ages per sample used in this study are provided in Table 1, with detailed U-Pb results given in Table DR1.1

Analytical Methods

Zircon crystals were extracted from sedi-ment samples by traditional methods of heavy mineral separation at the Arizona LaserChron

30°N

40°130°E

120°

Chinese Loess PlateauSand desertModern dust-storm trackGlacial dust-storm track

AQilian Shan

Altai

Qinling Shan

Tianshan

AltunTagh

Pamir

YellowRiver

Tibet

Kunlun Shan

QaidamBasin

BC

GobiDesert

Taklimakan Desert

500 km

Figure 1. Schematic map of Central Asia, showing the modern, simplified dust-storm tracks and near-surface wind patterns (red arrows), through-out the (A) Mu Us, (B) Tengger, and (C) Badan Jaran Deserts, and the hypothetical “glacial” dust-storm track (blue arrows), through the Qaidam Basin and across northern and central Tibet (Pullen et al., 2011).

1GSA Data Repository item 2016026, U-Pb age distributions of samples, is available at http:// www .geosociety .org /pubs /ft2016 .htm or by request to editing@ geosociety .org.

http://www.geosociety.org/pubs/ft2016.htm

http://www.geosociety.org/pubs/ft2016.htm

mailto:editing%40geosociety.org?subject=

Licht et al.


center (e.g., Gehrels et al., 2011). To mini-mize the hydraulic sorting grain-size effects and thus loss of fine-grained zircons in eolian dust samples, we used the recently developed ultrasonic separation technique of Hoke et al. (2014). Isotope ratios were measured using a Nu Plasma high-resolution (HR) multicollec-tor–inductively coupled plasma–mass spec-trometer (MC-ICP-MS) and a Thermo Element 2 single-collector (SC) ICP-MS coupled to a Photon Machines Analyte G2 193 nm excimer laser. Coarser-grained samples from the dust source areas were ablated using a 30 mm laser beam diameter at 7.0 mJ (constant energy) with 94% laser energy at 7 Hz for the Nu Plasma and 8 Hz for the Element 2. Finer-grained samples (i.e., loess-paleosol) were analyzed using a 12 mm beam diameter with same laser fluence and were analyzed on the Nu Plasma. Initial Pb was corrected using the 204Pb measurement and assuming an initial Pb composition from Stacey and Kramers (1975; see Table DR1 for more detail [footnote 1]). Note that the analyzed dust zircons have a minimum diameter of 12 µm due to ICP-MS collector limitations, ion yields, and the Pb and U concentrations of the zircons. This resolution is nevertheless better than for previous eolian dust provenance studies, which commonly used laser spots >25 µm (e.g., Che and Li, 2013; Nie et al., 2014). Accordingly, the results presented in this study are valid for the eolian dust coarse fraction only. The ori-gin of the finer fraction is beyond the focus of this study but has been discussed elsewhere (e.g., Derbyshire et al., 1998; Sun et al., 2006; Stevens and Lu, 2009).

Assessing Sample-Size Effects

Provenance studies based on U-Pb age distri-butions of detrital zircons usually aim to iden-tify all age components in samples of unknown provenance and compare them with the age components of known sources (Gehrels, 2000). To that end, 100–300 zircon ages per sample are commonly considered as a reasonable range to minimize the probability of failing to identify age components at low analytical cost (Ver-meesch, 2004; Andersen, 2005).

The provenance problem that we face here is different, because all of the dominant age com-ponents are to some extent shared by the known eolian sources. Age distributions of the poten-tial sources regions mainly differ in the relative contribution of four major age components: 225–330 Ma; 380–500 Ma; 750–1000 Ma; and 1500–2800 Ma (e.g., Gehrels et al., 2003; Stevens et al., 2010; Pullen et al., 2011; Lease et al., 2007, 2012; Weislogel et al., 2010). Previ-ous provenance studies have attempted to distin-guish among these contributions by comparing the relative magnitude of these age peaks in the age distribution of loess samples (Xiao et al., 2012; Nie et al., 2014). Although n = 100–300 ages per sample are sufficient to ensure that each age component is found in the age distribu-tions, these numbers are too low to constrain the relative contribution of each age peak (Pullen et al., 2014). As noted by Pullen et al. (2014), there is no minimum fixed number n that would completely stabilize the shape of the age dis-tributions, but rather a range (n = 300–1000), depending on the number of age groups present.

A good example of the uncertainty introduced by small n is shown in Figure 3. Subsamples made of randomly picked n = 100 ages from a bigger sample of n = 928 ages yield peak magnitudes that are different from each other as well as that of the larger-n age population. Previous loess provenance studies comparing age peak mag-nitudes used samples with n commonly <300 and have thus overinterpreted peak magnitude differences as real source differences, when in fact they may simply be due to scatter from the effects of subsampling. The same subsampling uncertainty also affects the MDS statistical tech-nique of Vermeesch (2013) when applied to the loess provenance problem. The MDS produces a “map” of points, representing samples, on which the “distance” between points is related to the dissimilarity between samples, quanti-fied by the KS statistic (Vermeesch and Gar-zanti, 2015). However, the KS statistic between two samples is based on the difference in their empirical distribution functions, the shape of which is controlled by subsampling as well (Massey, 1951). This uncertainty is illustrated in Figure 4. On the MDS plot, the random sub-samples are widely scattered around the bigger sample, with some of them falling next to the Qaidam Basin age compilation, whereas others do not. Though MDS plots are an elegant way to visualize group differences when dealing with a huge number of samples, the distance between individual samples is influenced by subsam-pling variability and does not allow a quantita-tive estimate of the mismatch.

The age distributions of the 23 new indi-vidual samples are plotted in Figure 5. For all of them, main age peaks fit within the range of age peaks previously mentioned to be found in central China. However, the amplitude of age peaks in these individual samples, with n between 70 and 370, is expected to approxi-mate but not perfectly characterize the “true” distribution of ages because of subsampling effects. To that end, the visualization and analy sis of each sample age distribution are thus insufficient for solving our provenance problem. To minimize the uncertainty intro-duced by subsampling, our first step was to group individual samples into geographical provinces such that each province had a suf-ficiently large n number to increase the statisti-cal significance of each age peak contribution (n between 800 and 2500). We divided the potential dust-source regions into four prov-enance provinces (Fig. 2):

(1) the sandy Mu Us Desert, located directly to the north of the Central Loess Plateau;

(2) the central sand deserts, including both the Tengger and Badan Jaran Deserts, located to the northwest of the Chinese Loess Plateau,

Plio-Quaternary depositsLoess Plateau deposits

Samples ( : This study)( )Mu Us DesertCentral sand deserts

Qaidam BasinYellow RiverLoess (& paleosols)Tibetan Cenozoic volcanics

N

Qaidam B nasi

Qilian Shan

TarimBasin

Centralsand deserts

Mu UsDesert

Xi’an

Beijing

80°E 90°E 100°E 110°E

30°N

40°N

500 km

GobiDesert

Altai

Kunlun Shan

TD

BJD

HS

Yellow River

Figure 2. Map of central China displaying the four main potential source provinces for coarse eolian dust and the samples used in this study (nature of the samples detailed in Table 1). TD—Tengger Desert; BJD—Badan Jaran Desert; HS—Heimugou Section.



TABLE 1. NATURE, LOCATION, AND NUMBER OF U-Pb AGES FOR THE SAMPLES USED IN THE STUDY

Sample name n NatureCoordinates

Reference(°N) (°E)Mu Us Desert13DUNE01 94 Eolian dune, Central Ordos Basin 38.98389 108.16321 This paper13COSMO01 84 Top of yardang (Mesozoic sandstone) 38.9746 107.3386 This paper13COSMO02 92 Top of yardang (Mesozoic sandstone) 38.972 107.343 This paper13COSMO03 90 Top of yardang (Mesozoic sandstone) 39.1011 107.8649 This paperMD10 98 Eolian dune, western Mu Us 37.9239 107.991 Stevens et al. (2013)MD09 108 Eolian dune, western Mu Us 38.4923 107.227 Stevens et al. (2013)MD08 119 Mesozoic eolian sandstone, eastern Mu Us 39.573 108.5151 Stevens et al. (2013)MD07 93 Eolian dune, eastern Mu Us 38.6528 109.6644 Stevens et al. (2013)MD06 58 Mesozoic eolian sandstone, eastern Mu Us 37.992 108.87 Stevens et al. (2013)MD05 115 Quaternary fluvial deposits, eastern Mu Us 37.9871 108.822 Stevens et al. (2013)MD04 74 Eolian dune, western Mu Us 37.706 108.488 Stevens et al. (2013)MD03 113 Quaternary sand dune, western Mu Us 37.722 108.488 Stevens et al. (2013)MD02 117 Eolian dune, eastern Mu Us 38.477 108.763 Stevens et al. (2013)MD01 152 Eolian dune, eastern Mu Us 38.136 109.791 Stevens et al. (2010)

Central deserts13DUNE03 81 Eolian dune, Tengger Desert 40.07233 103.92398 This paper13YARD01 82 Top of yardang, Badan Jaran Desert 40.5694 96.7546 This paper13DUNE05 88 Eolian dune, Tengger Desert 39.28238 102.69492 This paper13DUNE06 91 Eolian dune, Tengger Desert 38.80672 102.36845 This paper13YARD02 66 Top of yardang, Badan Jaran Desert 40.48549 94.63945 This paperTD1 86 Eolian dune, Tengger Desert 38.589 105.478 Stevens et al. (2010)YG02 235 Fluvial sandstone, foothills of the Gobi-Altai Mountains 41.7 103.1 Che and Li (2013)RSH01 230 Fluvial sandstone, Ruoshi River 40.4 99.6 Che and Li (2013)13DUNE02* 82 Eolian dune, Tengger Desert 39.37441 106.5782 This paper*13DUNE07* 94 Eolian dune, Yardang National Park 40.53599 93.10771 This paper*13DUNE09* 90 Eolian dune, Yardang National Park 39.71717 94.35859 This paper*13COSMO07* 84 Top of yardang (Pliocene–Quaternary sandstone) 40.53422 93.10758 This paper*13COSMO09* 83 Top of yardang (Pliocene–Quaternary sandstone) 40.47707 93.13603 This paper*

Qaidam Basin10QBAP04 88 Quaternary alluvium 38.3 93.416667 Pullen et al. (2011)MT226DZ6090 90 Pliocene–Quaternary lacustrine 38.3 93.416667 Pullen et al. (2011)MT226DZ4060 92 Pliocene–Quaternary lacustrine 38.3 93.416667 Pullen et al. (2011)MT226DZ0040 86 Pliocene–Quaternary lacustrine 38.3 93.416667 Pullen et al. (2011)MT_97DZ4060 93 Pliocene–Quaternary lacustrine 38.3 93.416667 Pullen et al. (2011)MT_97D2040 93 Pliocene–Quaternary lacustrine 38.3 93.416667 Pullen et al. (2011)4–18–09–1 86 Top of yardang 38.2428 92.4412 This paper (see also Rohrmann et al., 2013)4–17–09–1 90 Top of yardang 37.9667 92.7961 This paper (see also Rohrmann et al., 2013)4–27–09–1 161 Top of yardang 38.3582 92.5104 This paper (see also Rohrmann et al., 2013)

Northeast Tibet, Yellow River13YELLOW01 367 River sediment 39.40816 106.7366 This paperYR-1 121 River sediment 36.1388 103.609 Stevens et al. (2013)CH12–21 118 River sediment 35.879167 102.22583 Nie et al. (2015)YR-1 121 River sediment 36.139483 103.60908 Nie et al. (2015)CH11YR03–04 321 River sediment 37.521967 105.66667 Nie et al. (2015)YR-9C 108 River sediment 36.081139 103.86439 Nie et al. (2015)

Loess deposits—Chinese Loess PlateauHL1 84 Loess, Heimugou section 35.725883 109.06631 Pullen et al. (2011)HL9 165 Loess, Heimugou section 35.725883 109.06631 Pullen et al. (2011)HL15 239 Loess, Heimugou section 35.725883 109.06631 Pullen et al. (2011)HL33 39 Loess, Heimugou section 35.725883 109.06631 Pullen et al. (2011)XF-L1 90 Loess, Xifeng section 35.883 107.966 Xiao et al. (2012)10YG-1 96 Loess, Weinan section 34.35 109.516 Xiao et al. (2012)ljg 131 Loess, Xifeng section 35.78 107.78 Che and Li (2013)LGM Loess 84 Loess, Huanxian section 36.6224 107.2867 Stevens et al. (2010)

Paleosols—Chinese Loess PlateauXF-S1 68 Paleosol, Xifeng section 35.883 107.966 Xiao et al. (2012)10YG-2 88 Paleosol, Weinan section 34.35 109.516 Xiao et al. (2012)TJZ-1 81 Paleosol, Xining section 36.45 101.783 Xiao et al. (2012)slgj 261 Paleosol, Xifeng section 35.78 107.78 Che and Li (2013)S-0 94 Paleosol, Heimugou section 35.725883 109.06631 This paperS-1 108 Paleosol, Heimugou section 35.725883 109.06631 This paperS-9 103 Paleosol, Heimugou section 35.725883 109.06631 This paperS-15 105 Paleosol, Heimugou section 35.725883 109.06631 This paperS-22 95 Paleosol, Heimugou section 35.725883 109.06631 This paper

Note: Global positioning system coordinates for samples from other studies, when not specified in their original publication, were estimated from published maps.*Samples have not been included in the regional clusters because either their location is at the interface between two clusters (sample 13DUNE02), or because they

are located too far from the major dust-storm tracks supplying the Chinese Loess Plateau (see Figs. 1 and 2).

Licht et al.


extending from the Qilian Mountains in the south to the Altai Mountains in the north;

(3) the Qaidam Basin, located to the west of the Chinese Loess Plateau and bounded by the Nan-Qilian, Altyn, and Kunlun mountains to the northeast, northwest, and south, respec-tively; and

(4) the Upper Yellow River drainage system in northeast Tibet, which transports sediment to the north of the Chinese Loess Plateau that can

be reworked by wind (Stevens et al., 2013; Nie et al., 2015).

These four provinces are considered to be the only direct providers of coarse-grained dust. A contribution from the Taklimakan and Gobi Deserts, located further west and north, has also been proposed (Che and Li, 2013; Nie et al., 2014; Yan et al., 2014). However, we argue that any coarse (>12 µm) eolian zircon supply from these regions would have to transit via saltation

across the Mu Us, central sand deserts, and/or the Qaidam Basin and that its signature should be represented in the zircons of these three regions (Amit et al., 2014).

For the four selected source province regions, we compiled previously published and our new U-Pb zircon ages from eolian, fluvial, and lacustrine deposits, as well as from wind-eroded strata (see Table 1 for data compilation). All the selected U-Pb ages come either from Plio-cene–Quaternary deposits, or from older strata (Miocene to Cretaceous) that exhibit evidence of having been wind eroded.

Note that five new samples were not incor-porated in these four regional clusters (Table 1), because they are located either too far from the Chinese Loess Plateau, in areas character-ized by easterly surface winds blowing into the Taklimakan (samples 13DUNE07, 13DUNE09, 13COSMO07, and 13COSMO09), or at the geographical boundary between two prove-nance provinces (sample 13DUNE02).

We divided the dust deposits of the Chinese Loess Plateau into two main groups:

(1) the Quaternary loess sequences, com-monly numbered from L1 to L33 (younger to older), which were deposited during glacial periods (An et al., 1990); and

(2) the paleosol layers, numbered from S0 to S32, which experienced pedogenesis during interglacial periods (An et al., 1990; Porter and An, 1995).

For these two groups, we compiled new and existing U-Pb ages from samples located on the south and eastern parts of the Chinese Loess Plateau, excluding localities that are proximal to the potential provenance provinces, in order to avoid local bias in supply (i.e., Che and Li, 2013; Xiao et al., 2012).

Figure 3. Kernel density esti-mators (KDE) and histograms (20 Ma bins) for the loess layers of the Chinese Loess Plateau (CLP; n = 928; see Re-sults section) and five random sub samples (n = 100), shown for the interval 0–1500 Ma. The KDE and histograms for the Qaidam Basin (n = 879) are shown for comparison. The kernel density bandwidth of each plot was determined with plug-in bandwidth selection method of Botev et al. (2010). Whereas the 380–500 Ma age peak is clearly distinct and slightly higher than the 225–330 Ma age peak in Chinese Loess Plateau total age distri-bution, this pattern can be lost in smaller subsamples (peaks of similar size in subsample b; peaks nonexistent in sub samples c and d) or significantly increased (380–500 Ma age peak prominent in subsample e, as seen in Qaidam Basin age distribution). This illustrates the uncertainty introduced by small-n samples in interpreting relative age peak magnitudes.

CLP ( , n=928Loess layers)

CLP subsample a, n=100

CLP subsample b, n=100

CLP subsample c, n=100

CLP subsample d, n=100

CLP subsample e, n=100

0 500 1000 1500

Qaidam Basin, n=879

0 500 1000 1500Detrital zircon age (Ma)

Qaidam Basinn=879

CLPtotal

CLP aCLP b

CLP c

CLP dCLP e

–0.1 0.10

0

–0.1

0.1

Dim 1

Dim2

Figure 4. Multidimensional scaling (MDS) map for the loess lay-ers of the Chinese Loess Plateau (CLP total), the five random sub-samples of Figure 2, and Qaidam Basin. Axes are in dimensionless “KeS units” (0 < Kolmogorov-Smirnov [KS] < 1) of dissimilarity between samples (Vermeesch, 2013). Solid lines and dashed lines connect samples with their “closest” and second-closest neighbors, respectively. Final “stress” value is 0.098, indicating a fit between “fair” and “good” (Vermeesch, 2013). Note that subsamples are ran-domly scattered around CLP total, with distances similar to the one separating Qaidam Basin and Chinese Loess Plateau age distribu-tions. The uncertainty introduced here by subsampling does not al-low these two areas to be distinguished.



Statistical Analysis

The second step of our approach was to focus on the age distributions of the six geographical regions (the four provenance provinces and the two dust groups). The amplitudes of the age peak contributions are meaningful here given the large number of zircon ages (n > 800) deter-mined for each region. The age distributions can be visually compared using histogram dia-grams, probability density plots (PDPs), and kernel density estimators (KDEs; Vermeesch, 2012). PDPs are omitted from the figures in this paper because KDE and PDP converge toward a single representation of the distribution of ages for large-n data sets (Pullen et al., 2014). In an attempt to quantify sample age components and their contribution, we used the freeware Bayes-Mix (Compston and Gallagher, 2012), which statistically determines the number, age, width, and contribution of individual components in age distributions. BayesMix uses a reversible jump Markov chain Monte-Carlo (RJ-MCMC) approach, which is an iterative, Bayesian sam-pling strategy that allows the number of indi-vidual components to change, as well as their age, width, and contribution, in order to best fit a given age distribution (Jasra et al., 2006; Gal-

lagher et al., 2009). For each age data set used in the paper, BayesMix was run for 15,000 iter-ations (details in Table DR2 [see footnote 1]), assuming heavy skew-t distributions with equal weighting on skewness.

One of the most promising attempts to quan-titatively determine the contribution of different sources with complex age distributions in sam-ples of unknown sourcing was provided by the iterative mixture modeling strategy of Amidon et al. (2005), and which was subsequently used on Tibetan samples (Lease et al., 2007, 2012). The iterative approach for mixture modeling of Amidon et al. (2005) consists of modeling synthetic age distributions by proportionally combining the probability density functions (PDFs) of the potential sources for the whole range of possible source combinations. Then, the best mixture is found by minimizing the dis-similarity between the synthetic PDFs and the PDF of the sample (or sets of samples in this case) of unknown sourcing, using the KS sta-tistic or the areal percent mismatch as dissimi-larity measures. This approach implies that the PDFs of each potential source region perfectly reflect the whole range of ages that contributes to the mixture. Erosion and sediment transport are nonstationary processes, however, and only

a small, random subsample of zircons from each source reaches the point of deposition. To take into account this potential variability in the zir-con supply and the potential uncertainty intro-duced by random subsampling during erosion and transport, we developed a modified version of the strategy used by Amidon et al. (2005).

For each possible combination of the four identified source regions (for instance, Mu Us Desert: 20%, central sand deserts: 30%, Qaidam Basin: 12%, and northeast Tibet: 38%), we modeled N synthetic age distributions of 800 ages by randomly picking the necessary amount of ages in each source data set. N was chosen to be large enough to reproduce the potential variability in zircon subsampling. We then cal-culated the dissimilarity between each of these N synthetic age distributions and the loess (or paleosol) age distribution. There are different ways to determine the combinations that best fit the loess (or paleosol). A first approach is to determine the minimum dissimilarity value dmin among the N synthetic age distributions and for the whole array of potential source combina-tions. The combinations that best fit the loess (or paleosol) are those for which this minimum dmin value is reached for at least one of their N synthetic distributions. A second approach

S-9, =103n

S-15, =105n

S-22, =95n

13YELLOW01, =367n

13DUNE01, =94n

13DUNE02, =82n

13DUNE03, =81n

13YARD01, =82n 13DUNE05, =88n

13DUNE06, =91n13YARD02, =66n

13DUNE09, =90n

13COSMO01, =84n

13COSMO02, =92n

13COSMO03, =90n

13DUNE07, =94n

13COSMO07, =84n

13COSMO09, =83n

4-18-09-1, =86n

4-17-09-1, =90n

4-27-09-1, =161n

S-0, =94n

S-1, =108n

N90°E 100°E 110°E

35°N

40°N

500 km

1000 2000 30000Detrital zircon ages (Ma)




1000 2000 30000Detrital zircon ages (Ma) 1000 2000 30000

Detrital zircon ages (Ma)

Figure 5. Kernel density plots of the 23 samples presented in this study. The kernel density bandwidth of each plot was determined with plug-in bandwidth selection method of Botev et al. (2010). Same legend as Figure 2.

Licht et al.


is to consider the average dissimilarity value among the N synthetic age distributions per potential source combination. The combina-tions that best fit the loess (or paleosol) are those for which this average dissimilarity value is the lowest (named here Dmin). This way of determining the best fit takes into account the reproducibility of the dissimilarity values per source combination but minimizes the poten-tial effect of subsampling in the contribution of the source provinces. A final approach is to consider the range of source combinations for which more than half (or two thirds, depend-ing on the required precision) of the N synthetic distributions are “statistically similar” to the loess (or paleosol) age distribution. The notion of “statistical similarity” depends on the chosen dissimilarity measure.

We used the KS statistic as the dissimilarity measure, which has been shown to be an effec-tive dissimilarity indicator (Vermeesch, 2013). We also tried another dissimilarity measure, the Anderson-Darling statistic, and it yielded no significant impact on the results (not shown). N = 200 was recognized to be sufficient to stabilize the average dissimilarity values and their standard deviation. Two age distributions are considered statistically identical when the two-sample KS statistic does not reject the null hypothesis at the 95% confidence level.

RESULTS AND INTERPRETATION

U-Pb Age Distributions

Probability density plots, age histograms, and main age components for the maximum posterior models determined by the RJ-MCMC approach for the six geographic provinces are displayed in Figure 6. Mean, mode, standard deviation, and skew parameters for all compo-nents are given in Table DR2 (see footnote 1). Note that age components in the maximum posterior models are overall Gaussian, and that only the old (older than 1500 Ma) components display skewed distributions. Cumulative prob-ability plots are shown in Figure 7.

The age distributions of the four potential provenance regions and loess and paleosol layers share several similar features. Grains younger than 700 Ma contribute from 40% (in the Mu Us Desert) to 60% (Qaidam Basin, northeast Tibet), and up to 70% (in the central sand deserts) of the total populations (Fig. 7). These young grains are divided into two major populations, centered at 240–270 Ma and 440–480 Ma, with relatively narrow distributions (1s < 20 Ma). Older grains are grouped into three main populations with wider distributions (1s commonly > 50 Ma), centered at 750–1130 Ma,

at 1645–1980 Ma, and 2430–2640 Ma, contri-butions of which vary from 2% to 25% of the total distributions.

Several distinctions can be made between the northern (Mu Us and central sand deserts) and southwestern provinces (Qaidam Basin and Upper Yellow River drainage system), as well as among individual regions:

(1) In the two northern provinces, the 240–270 Ma age peak is prominent (28%–48% of the total grains) and contributes twice as much as the 440–480 Ma age peak to the total age popu-lation. In contrast, the 440–480 Ma age peak is prominent in the southwestern provinces (30%–46% of the total grains). The 440–480 Ma age peak contributes ~25% more than the 240–270 Ma age peak to the total age population in Upper Yellow River sediment, and almost three times more in the Qaidam Basin.

(2) The zircon population centered at 750–1130 Ma is particularly important in the Qaidam Basin (~20%) and in Upper Yellow River sedi-ment (~16%), and almost nonexistent in the northern provinces (2%–5%).

(3) The Mu Us Desert is distinguishable in having a large contribution (~60%) of grains older than 1500 Ma, whereas this contribution is less in other regions (~30%).

The age distributions of Yellow River, loess, and paleosol layers do not show any significant difference, and the contributions of major age peaks vary by only 2%–5% between the three distributions. The two-sample KS test between loess and paleosol layers yields a high P value (P = 0.86), indicating that both distributions are statistically indistinguishable; this P-value is lower between Yellow River and loess (P = 0.18) and between Yellow River and paleosol layers (P = 0.65), though still indicating statisti-cal similarity between age distributions. Grains younger than 700 Ma contribute ~50% of the total population; the 440–480 Ma age peak con-tributes 17%–32% more than the 240–270 Ma age peak to the total age population. The zircon population centered at 750–1130 Ma is impor-tant and contributes between 11% and 16%. Zir-cons older than 1500 Ma constitute the remain-ing 30% of the total age population. Note that Yellow River, loess, and paleosol layers yield a minor (1%–4%) population of very young grains (younger than 65 Ma), already noted by Pullen et al. (2011), that is absent in the three other potential provenance regions.

Qualitative Assessment of Eolian Dust Provenance

The prominence of the 440–480 Ma age pop-ulation in eolian dust samples compared to the 240–270 Ma age population (17%–32% higher)

is consistent with Upper Yellow River and/or Qaidam Basin being dominant sources. Statisti-cal and visual similarity between loess or paleo-sol layers and Yellow River sediment indicates that the latter is likely the prominent source. However, minor contributions of coarse-grain zircon crystals from other provinces cannot be rejected if they are combined. Contributions of Upper Yellow River main age peaks appear to be intermediate between those of the Qaidam Basin and those of the two northern provinces. A combined contribution of both these regions could potentially result in a mixed age distri-bution that is similar to those of Yellow River sediment. We conclude that the eolian dust age distributions reflect a prominent contribution from the Yellow River, with a potential, minor additional supply from the three other provinces that is challenging to assess visually.

Quantitative Assessment of Eolian Dust Provenance

We used the mixture modeling strategy intro-duced in Methods section to determine the com-bination of the four potential source regions that best fits the age distributions of loess and paleosol layers. The average dissimilarity to the loess and paleosol layers is displayed in ternary diagrams spanning the whole possible array of source combinations (Figs. 8 and 9). Combina-tions yielding the minimum dissimilarity value dmin and lowest average dissimilarity value Dmin are given in Table 2.

For loess layers, lowest average dissimilarity values (0.04–0.06) were obtained for a contri-bution of Qaidam Basin between 0% and 30%, contribution of the two northern provinces between 0% and 20% each, and a contribution of Yellow River sediment between 40% and 100% (Fig. 8). The range of combinations for which more than two thirds of the N random synthetic distributions are statistically similar to the loess is broader, with a Yellow River contribution that can decrease to 10%, mainly counterbalanced by an increasing contribution from Qaidam Basin. The minimum dissimilar-ity value dmin and the lowest average dissimi-larity Dmin are relatively similar and indicate a best fit for a common contribution from the Yellow River of 58%–68%, a contribution from the Qaidam Basin between 14% and 20%, and the two northern provinces contributing the remainder. P values of the KS statistic for these combinations are >0.8, indicating a very good fit. These P values are higher than those for the KS test between Yellow River and loess age dis-tributions.

The combinations that best fit the age distri-bution of paleosol layers are similar to those for



the loess layers (Fig. 9). This is not surprising, considering the visual similarity between paleo-sol and loess age distributions. The minimum dissimilarity value dmin indicates a best fit for a contribution from the Yellow River of 66%, a

contribution from the Qaidam Basin of 26%, and the two northern provinces contributing the remainder. The lowest average dissimilarity Dmin is slightly different, indicating a best fit for a single provenance from the Yellow River (96%

of the total supply); however, other low average dissimilarity values (0.04–0.06) occupy a broad range similar to the one for loess layers. P val-ues of the KS statistic for these combinations are also >0.8, indicating a very good fit.

µσ= 457 Ma= 5 Mac= 40 %

µσ= 258 Ma= 23 Mac= 13 %

µσ= 1646 Ma= 371 Mac= 17 %

µσ= 458 Ma= 52 Mac= 6 %

Qaidam Basin, =829n

1000 2000 30000

0.02

0.01

0

1000 2000 30000

0.02

0.01


µσ= 269 Ma= 7 Mac= 28 %

µσ= 458 Ma= 17 Mac= 13 %

µσ= 1905 Ma= 68 Mac= 19 %

µσ= 2588 Ma= 40 Mac= 22 %

10%

µσ= 272 Ma= 17 Mac= 46 %

µσ= 448 Ma= 28 Mac= 24 %

µσ= 1728 Ma= 308 Mac= 15 %

µσ= 261 Ma= 27 Mac= 24 %

µσ= 444 Ma= 18 Mac= 28 %

µσ= 834 Ma= 68 Mac= 11 %

µσ= 1580 Ma= 444 Mac= 21 %

µσ= 263 Ma= 32 Mac= 22 %

µσ= 478 Ma= 24 Mac= 29 %

µσ= 842 Ma= 167 Mac= 16 %

µσ= 1984 Ma= 306 Mac= 24 %


Probability

Probability

contribution(c)

10%

contribution(c)

10%

contribution(c)

Central sand deserts, =959nMu Us Desert, =1407n

Loess layers, CLP =928n Paleosol layers, CLP = 1003n

Probability

1000 2000 30000

1000 2000 300001000 2000 30000

0

0.02

0.01

0

µσ= 255 Ma= 26 Mac= 24 %

µσ= 448 Ma= 22 Mac= 30 %

µσ= 1818 Ma= 146 Mac= 24 %

µσ= 910 Ma= 125 Mac= 16 %

Upper Yellow River, =1156n

1000 2000 30000

0.02

0.01

0

0.02

0.01

0

0.02

0.01

0

Figure 6. Normalized kernel density plots, histograms (40 Ma bins), and main components for the four potential source provinces (Mu Us Desert, central sand deserts, Qaidam Basin, and northeast Tibet), and the loess and paleosols layers of the Chinese Loess Plateau (CLP), shown for the interval 0–3500 Ma. Main components were obtained using the software BayesMix (see Methods section). The height of each vertical line reflects the component contribution (c) to the age population. Mean (μ) and standard deviation (σ) are given for components with contribution >10%. Note the juxtaposition of two components with similar mean age (457–458 Ma) but different standard deviations for the Qaidam Basin. Mean, mode, standard deviation, and skew parameters for all components are given in Table DR2 (see text foot-note 1). Old (older than 1500 Ma) age peaks appear to be better resolved when slightly skewed in BayesMix maximum posterior models, explaining the visual offset between age peaks in density plots and components in several samples (peak at 2588 Ma for the Mu Us Desert).

Licht et al.


DISCUSSION

Eolian Sedimentary Budget on the Chinese Loess Plateau

Both visual and statistical assessments of eolian dust age distributions suggest that the Yellow River is the dominant source (>50%) for the coarse fraction of the eolian deposits of the Chinese Loess Plateau. Our statistical approach

also shows that dust age distributions are bet-ter explained with an additional, combined sup-ply from the three other provinces, but not as significant as previously proposed (An et al., 1990; Ding et al., 1999; Pullen et al., 2011; Amit et al., 2014).

Our approach included mixing U-Pb data from various locations on the Chinese Loess Plateau to build large-n eolian dust U-Pb data sets. Thus, our analysis cannot address regional

variability in sediment sourcing on the loess plateau, which would be accomplished through analysis of single sample large-n data. In that sense, our results do not contradict previous U-Pb studies emphasizing the importance of the Mu Us Desert and the Yellow River as a major eolian source in the northern part of the Chinese Loess Plateau (Stevens et al., 2013; Bird et al., 2015; Nie et al., 2015), or northeast Tibet in its westernmost parts (Xiao et al., 2012; Che and Li, 2013; Bird et al., 2015). However, we argue that most of all the interpretations from previ-ous U-Pb studies should be taken with caution because they are based on only 80–200 ages per sample (i.e., Stevens et al., 2010, 2013; Pullen et al., 2011; Xiao et al., 2012; Che and Li, 2013; Nie et al., 2014), and may thus be influenced by age subsampling effects (e.g., Fig. 3). Simi-larly, grouping U-Pb ages in regional clusters results in a loss of information on variation in the sediment composition in the source regions themselves. For instance, previous studies have emphasized variability of U-Pb age distributions within the Mu Us Desert, with western Mu Us sands showing a significant contribution from eolian recycling from Yellow River sediment (Stevens et al., 2010, 2013). Localized sourc-ing from more spatially limited areas cannot be

1000 2000 30000

0

0.2

0.4

0.6

0.8


Cumulativeprobability

Qaidam BasinCentral sand desertsMu Us Desert

Loess layers, CLPPaleosol layers, CLP

Yellow River

Figure 7. Cumulative probabil-ity density plots for the four po-tential source provinces (Mu Us Desert, central sand deserts, Qaidam Basin, and upper Yellow River), and the loess and paleosols layers of the Chinese Loess Plateau (CLP), shown for the interval 0–3500 Ma.

C = 90%

0.1 0.2

Dissimilarity(Average Kolmogorov-Smirnov statistic)

Statistically identical at 95 % confidence in morethan 2/3 of the randomly generated mixtures

AB

D

D

D

D

D

D

D

D

D

D

AA

AA

AA

AA

A

B

C = 80%

C = 70%

C = 60%

C = 50%

C = 40%

C = 30%

C = 20%

C = 10%

C = 0%

10 30 50 70

7050

3010

7050

3010

B(centralsanddeserts)

A(Mu UsDesert)

D(northeastTibet)

C (Qaidam Basin) = 20 %for

0

Figure 8. Dissimilarity to the loess layers of the Chinese Loess Plateau for each possible combination of the four potential source regions (A—Mu Us Desert; B—central sand deserts; C—Qaidam Basin; D—upper Yellow River). Each triangle is a ternary diagram of contribu-tion from the provinces A, B, and D for a given contribution of C (contributions of A + B + C + D = 100% for each triangle; see example for C = 20% on the left side of the figure). The color bar indicates the range of values for the dissimilarity measure, here the Kolmogorov-Smirnov (KS) statistic averaged for N random synthetic distributions per source combination (N = 200). The combinations that best fit the loess age distribution are the ones that give the lowest dissimilarity values. An alternative way to determine the best fit is to look at the range of combinations for which more than two thirds of the N random synthetic distributions are statistically similar to the loess (here in the sense of the KS statistic at the 95% confidence level), depicted by red isolines in the figure.



investigated with our methodology and current data set, but it is likely that the continuously increasing number of U-Pb data from central China will allow it in the near future. In this regard, our study is valuable because it provides an approach about how to assess these increas-ingly large-n data sets. Finally, it is important to remember that we only investigated a fraction of the provenance signal by using U-Pb ages of “coarse” zircons (laser spot diameter of 12 µm for dust samples; 30 µm otherwise). Relative contribution estimates may change significantly when smaller grain sizes are considered.

Our eolian budget estimates confirm that Yellow River input buffers the contribution of other sources to the Chinese Loess Plateau for the dust coarse fraction (Stevens et al., 2013; Nie et al., 2015). This prominent contribution is corroborated by heavy mineral counting, framework petrography, and sediment flux esti-mates (Nie et al., 2015). These results do not completely rule out other modes of transporta-tion for the Tibetan-sourced dust, because they cannot tell the difference between zircons that came from Tibet via the Yellow River and were then reworked by winds further north (Stevens et al., 2013; Nie et al., 2015), and zircons that were blown directly off the Yellow River drain-age basin and landed on the Chinese Loess Pla-teau (Pullen et al., 2011). They do, however, corroborate minor yet significant supply of coarse sediment from the Qaidam Basin (Pul-len et al., 2011; Kapp et al., 2011). The range of potential Qaidam Basin contribution that could best explain the dust age distributions is yet relatively broad—and thus uncertain—in our statistical estimates (Figs. 8 and 9). This

is not surprising, because Qaidam Basin and Yellow River age distributions are relatively similar, in contrast to the age distributions of the two northern provinces. This is partly due to the fact that parts of the Yellow River source tributaries drain northern Tibet and the Kunlun Mountains, areas that also bound and supply the Qaidam Basin (Yin et al., 2008). Given the importance of large n to distinguish both provinces, it is likely that their relative contri-butions to the eolian budget will continue to change as more data are generated. Our results yet corroborate the existence of an alterna-tive dust-storm pathway through the Qaidam Basin, infrequent today but that must have been active in the past (Vandenberghe et al., 2006; Kapp et al., 2011; Pullen et al., 2011; Rohrmann et al., 2013). Eolian erosion in the Qaidam Basin has been interpreted to be most significant during glacial periods (Kapp et al., 2011; Rohrmann et al., 2013). In this regard, a notable finding is the lack of a detectable difference in sourcing between glacial (loess) and interglacial (paleosols) deposits. Constant Yellow River contribution during both peri-ods is not surprising because fluvial rework-ing should have occurred regardless of most

changes in atmospheric circulation, as Yellow River fluvial sediment surrounds the Chinese Loess Plateau to the west and north (Nie et al., 2015). However, minor Qaidam Basin contri-bution to the eolian supply during interglacials is counterintuitive, because modern dust-storm outbreaks are very limited there (Sun et al., 2001). An explanation for this minor Qaidam component in paleosol layers is that part of the interglacial dust is also recycled from previ-ous glacial (and interglacial) deposits (Kohfeld and Harrison, 2003; van Loon, 2006; Kapp et al., 2015). Eolian dust recycling in loess accumulations has been shown to be a sig-nificant process in other areas (e.g., Mroczek, 2013; Sweeney and Mason, 2013). Extensive sediment reworking by wind would serve to homogenize the composition of glacial-aged loess and dust in interglacial-aged paleosols.

We thus suspect that the relative contribution of the northern and northwestern deserts com-pared to the Qaidam Basin was more important during interglacials, such as observed today (Wang et al., 2004), but this contribution has been significantly dampened by recycling of older loess. Similarly, the relative contribution from the Qaidam Basin during glacials was

A: Mu Us DesertB: Central sand desertsC: Qaidam BasinD: Northeast Tibet

0.1 0.2

C = 90%

AB

D

D

D

D

D

D

D

D

D

D

AA

A AA

AA A

A

B

C = 80%

C = 70%

C = 60%

C = 50%

C = 40%

C = 30%

C = 20%

C = 10%

C = 0%

Dissimilarity(Average Kolmogorov-Smirnov statistic)

0

Statistically identical at 95 % confidence in morethan 2/3 of the randomly generated mixtures

Figure 9. Dissimilarity to the paleosol layers of the Chinese Loess Plateau for each possible combination of the four potential source regions. Same legend as Figure 8 (A—Mu Us Desert; B—central sand deserts; C—Qaidam Basin; D—upper Yellow River).

TABLE 2. COMBINATIONS YIELDING MINIMUM DISSIMILARITY VALUE ∆MIN

AND LOWEST AVERAGE DISSIMILARITY VALUE δMIN

Loess layers Paleosol layersMinimum dissimilarity

value, δmin

(%)

Average dissimilarity value, ∆min

(%)

Minimum dissimilarity value, δmin

(%)

Average dissimilarity value, ∆min

(%)Mu Us Desert 8 6 6 0Central sand deserts 14 12 2 4Qaidam Basin 20 14 26 0Northeast Tibet 58 68 66 96

Licht et al.


likely greater than during interglacials, and this proportion has been reduced by reworking of older deposits.

“Eolian Cannibalism” and the Building of the Chinese Loess Plateau

Our contribution estimates imply that buffer-ing of dust supply by reworking of Yellow River sediment and intense reworking of older loess and paleosol deposits have served to mix eolian dust and homogenize their provenance signa-tures. This explanation is supported by similar heavy mineral assemblages for both loess and paleosol layers (Nie and Peng, 2014). It is also corroborated by recent geomorphological analy-ses of wind erosion features in the Mu Us Desert and on the northern part of the Chinese Loess Plateau, suggesting that the central and eastern Loess Plateau previously extended farther to the north and west of its modern windward escarp-ment, and that this extension has been reworked and redeposited further south during the late Quaternary (Kapp et al., 2015). Considering that eolian dust accumulation rates were commonly 2–20 times larger during glacial periods than dur-ing interglacial periods, with a high regional vari-ability (Kohfeld and Harrison, 2003; Sun et al., 2008), it is not surprising that recycling on the Chinese Loess Plateau homogenized eolian zir-con populations toward glacial conditions, with a significant Qaidam Basin (~20%) contribution.

These findings corroborate the idea that the Chinese Loess Plateau is a dynamic landform where deflation, lateral transport by winds and rivers, and aggradation are equally important (Nie et al., 2015; Kapp et al., 2015). Evidence of extensive reworking of fluvial and eolian sediment suggests that the extensive Quaternary loess sequences are just the tip of the iceberg of a much longer accumulation history. Older, pre-Quaternary loess deposits at the margin of the Chinese Loess Plateau are likely to have been cannibalized by winds, such as seen in the Mu Us Desert (Kapp et al., 2015), or reworked by fluvial processes and redeposited, leaving a few isolated remnant deposits (An et al., 2001; Guo et al., 2002; Alonso-Zarza et al., 2009; Licht et al., 2014).

Evidence of intense recycling and “over-representation” of Yellow River sediment and glacially sourced dust raise concerns about the relevance of several paleoclimatic proxies applied to Chinese loess sequences. Petrog-raphy, mineralogy, or geochemistry analy-ses of single-particle or bulk eolian sediment might be significantly either buffered by fluvial provenance and/or biased toward glacial dust, explaining why many studies found changes of sediment properties between loess and paleo-

sol layers to be much weaker compared to the regional variability (e.g., Gylesjö and Arnold, 2006; Jeong et al., 2011; Yan et al., 2014; Nie and Peng, 2014). We suggest that proxies linked to in situ pedogenic development (in contrast with proxies linked to basinwide pedo-genic development, as dust fine-fraction clay-minerology, likely related to weathering in the sediment source areas) are the only ones lack-ing bias toward glacial conditions and fluvial sediment reworking because they are indepen-dent of sediment sourcing. Notably, iron-oxide mineralogy, as well as magnetic susceptibility and other various magnetic parameters, have been shown to highlight significant differences between loess and paleosol layers (e.g., Maher and Thompson, 1991; Deng et al., 2004; Song et al., 2014). This can be explained by the over-all authigenic character of magnetic minerals, which form in situ in paleosol layers rather than being transported by winds (e.g., Maher and Thompson, 1991, and references therein). However, recycling of pedogenic microstruc-tures, including clay aggregates and iron-oxide nodules, has been shown to be significant in other loess accumulations (Mroczek, 2013). In that sense, paleoclimatic proxies linked to in situ pedogenic development might be biased to a small degree by the reworking of older Chinese paleosol layers as well, though this assumption is hard to verify.

Implications for the Genesis of the Chinese Loess Plateau

Our findings emphasize the prominent role of eolian reworking of freshly brought fluvial sedi-ment and older loess deposits in the building of the Chinese Loess Plateau. We thus suspect that two distinct phenomena might have controlled the recent Pliocene–Quaternary increase as well as the long-term Cenozoic increase in eolian accumulation rates (Sun et al., 2008, 2011).

First, increased denudation of northeast Tibet and consecutive fluvial transport further north must have significantly enhanced loess accumu-lation (Nie et al., 2015). Active deformation and resultant rock uplift have modified the topogra-phy and increased denudation of the northeast-ern margin of Tibet throughout the Cenozoic, with a recent acceleration since the middle–late Miocene (e.g., Métivier et al., 1998; Clark et al., 2010; Duvall et al., 2011; Li et al., 2011; Lease et al., 2012). Increased erosion due to the onset of oscillating glacial conditions (Peizhen et al., 2001) and the setup of the Yellow River in its modern course (Craddock et al., 2010; Nie et al., 2015) must have additionally increased fluvial sedimentary input into the Chinese Loess Plateau and its consecutive reworking by winds.

Moreover, we suspect that increased eolian accumulation rates might reflect enhanced recy-cling of older loess and fluvial strata as much as increased dust generation. Indeed, the long-term erosion of older eolian and fluvial strata and their recycling in younger eolian dust deposits should reduce the spatial extent and thickness of older loess deposits, whereas younger loess deposits would appear prominent in the fos-sil record. It is thus very likely that changes in accumulation rates or regional onsets of eolian sedimentation in the fossil record are strongly biased by this recycling effect. Note that we do not question the numerous other paleoclimatic proxy records demonstrating aridification and desert expansion since the late Miocene (e.g., Wang et al., 1999; Wang et al., 2006; Wu et al., 2007; Han et al., 2014). However, we suspect that increasing aridity throughout the Pliocene–Quaternary might have favored the reworking of older deposits on the Chinese Loess Plateau as much as dust generation in distant deserts.

CONCLUSION

Our study highlights how provenance inter-pretations based on U-Pb ages of detrital zircons can be misleading when based on a small num-ber of zircon ages. As already emphasized pre-viously (Pullen et al., 2014), we argue that inter-pretations based on age peak heights, as well as mixture modeling of provenance provinces, are only meaningful when the number of ages per region is high enough (here >800) to stabi-lize the shape of age distributions. We present a new mixture modeling method to quantify the contribution of different sources of known age distribution. Our method takes into account the potential uncertainties introduced by random zircon subsampling during erosion and transport and allows us to statistically estimate the com-bination of known sources that best fits the age distribution of samples of unknown sourcing.

Our contribution estimates based on available data indicate that the coarse eolian dust sedi-mentary budget on the Chinese Loess Plateau is dominated by Yellow River–brought sedi-ment (60%–70%). These results emphasize the importance of recycling of fluvial deposits in the loess sedimentary budget. Moreover, our find-ings confirm minor Qaidam Basin input (~20%) and thus support the existence of an alternative dust-generating storm pathway through the Qaidam Basin and northern Tibetan Plateau, likely active during past glacial and possibly stadial periods. In this regard, the lack of a sig-nificant difference in sourcing between glacial (loess) and interglacial (paleosols) deposits is particularly noteworthy. This implies that intense reworking of older loess and paleosol



deposits has additionally served to mix eolian dust and homogenize their provenance signa-tures. Our findings corroborate the idea that the Chinese Loess Plateau is a dynamic landform where wind deflation, lateral transport by wind and rivers, and aggradation are equally impor-tant. Moreover, they suggest that paleoclimatic interpretations based on loess accumulation rates or on eolian dust features might be sig-nificantly biased by these recycling effects. Finally, we propose that a significant part of the older, pre-Quaternary deposits surrounding the Chinese Loess Plateau might have been partly cannibalized and recycled in more recent loess deposits following increasing regional aridity throughout the Pliocene–Quaternary.

ACKNOWLEDGMENTS

This research was primarily funded by the U.S. Na-tional Science Foundation (NSF grants AGS-1203973 and AGS-1203427). Licht was also funded by a Fyssen Foundation study grant (2014), Kansas University (2015), and Marie Curie fellowship ECAMMETT 656731 (2015–2016). We also acknowledge NSF grant EAR 1338583 for support of the Arizona LaserChron Center. We thank H. Chang, N. Mclean, A. Möller, K. Gallagher, K.C. Beard, M. Pecha, G. Simpson, I. Nurmaya, and C. White for sample collection, pro-lific discussions, and assistance in the laboratory.

REFERENCES CITED

Alonso-Zarza, A., Zhao, Z., Song, C., Li, J., Zhang, J., Martin-Perez, A., Martín-García, R., Wang, X., Zhang, Y., and Zhang, M., 2009, Mudflat/distal fan and shal-low lake sedimentation (upper Vallesian–Turolian) in the Tianshui Basin, central China: Evidence against the late Miocene eolian loess: Sedimentary Geology, v. 222, p. 42–51, doi: 10 .1016 /j .sedgeo .2009 .03 .010 .

Amidon, W., Burbank, D., and Gehrels, G., 2005, U-Pb zircon ages as a sediment mixing tracer in the Nepal Himalaya: Earth and Planetary Science Letters, v. 235, p. 244–260, doi: 10 .1016 /j .epsl .2005 .03 .019 .

Amit, R., Enzel, Y., Mushkin, A., Gillespie, A., Batbaatar, J., Crouvi, O., Vandenberghe, J., and An, Z., 2014, Link-ing coarse silt production in Asian sand deserts and Quaternary accretion of the Chinese Loess Plateau: Geology, v. 42, p. 23–26, doi: 10 .1130 /G34857 .1 .

An, Z., Tunghseng, L., Yanchou, L., Porter, S.C., Kukla, G.H.W.X., Xihao, W., and Yingming, H., 1990, The long-term paleomonsoon variation recorded by the loess-paleosol sequence in central China: Quaternary International, v. 7, p. 91–95.

An, Z., Kutzbach, J., Prell, W., and Porter, S., 2001, Evolu-tion of Asian monsoons and phased uplift of the Hima-laya-Tibetan Plateau since late Miocene times: Nature, v. 411, p. 62–66, doi: 10 .1038 /35075035 .

Andersen, T., 2005, Detrital zircons as tracers of sedimen-tary provenance: Limiting conditions from statistics and numerical simulations: Chemical Geology, v. 216, p. 249–270, doi: 10 .1016 /j .chemgeo .2004 .11 .013 .

Bird, A., Stevens, T., Rittner, M., Vermeesch, P., Carter, A., Andò, S., Garzanti, E., Lu, H., Nie, J., Zeng, L., Zhang, H., and Xu, Z., 2015, Quaternary dust source variation across the Chinese Loess Plateau: Palaeogeography, Palaeoclimatology, Palaeoecology, v. 435, p. 254–264, doi: 10 .1016 /j .palaeo .2015 .06 .024 .

Botev, Z.I., Grotowski, J.F., and Kroese, D.P., 2010, Kernel density estimation via diffusion: Annals of Statistics, v. 38, p. 2916–2957, doi: 10 .1214 /10 -AOS799 .

Che, X., and Li, G., 2013, Binary sources of loess on the Chinese Loess Plateau revealed by U-Pb ages of zir-con: Quaternary Research, v. 80, p. 545–551, doi: 10 .1016 /j .yqres .2013 .05 .007 .

Clark, M.K., Farley, K.A., Zheng, D., Wang, Z., and Duvall, A.R., 2010, Early Cenozoic faulting of the northern Tibetan Plateau margin from apatite (U-Th)/He ages: Earth and Planetary Science Letters, v. 296, p. 78–88, doi: 10 .1016 /j .epsl .2010 .04 .051 .

Compston, W., and Gallagher, K., 2012, New SHRIMP zir-con ages from tuffs within the British Palaeozoic strato-types: Gondwana Research, v. 21, p. 719–727, doi: 10 .1016 /j .gr .2011 .11 .010 .

Craddock, W.H., Kirby, E., Harkins, N.W., Zhang, H., Shi, X., and Liu, J., 2010, Rapid fluvial incision along the Yellow River during headward basin integration: Nature Geoscience, v. 3, p. 209–213, doi: 10 .1038 /ngeo777 .

Deng, C., Zhu, R., Verosub, K.L., Singer, M.J., and Vidic, N.J., 2004, Mineral magnetic properties of loess/ paleosol couplets of the central loess plateau of China over the last 1.2 Myr: Journal of Geophysical Research–Solid Earth, v. 109, B01103, doi: 10 .1029 /2003JB002532 .

Derbyshire, E., Meng, X., and Kemp, R.A., 1998, Prov-enance, transport and characteristics of modern aeolian dust in western Gansu Province, China, and interpre-tation of the Quaternary loess record: Journal of Arid Environments, v. 39, p. 497–516, doi: 10 .1006 /jare .1997 .0369 .

Ding, Z., Sun, J., Rutter, N.W., Rokosh, D., and Liu, T., 1999, Changes in sand content of loess deposits along a north-south transect of the Chinese Loess Plateau and the implications for desert variations: Quaternary Research, v. 52, p. 56–62, doi: 10 .1006 /qres .1999 .2045 .

Duvall, A.R., Clark, M.K., van der Pluijm, B.A., and Li, C., 2011, Direct dating of Eocene reverse faulting in northeastern Tibet using Ar-dating of fault clays and low-temperature thermochronometry: Earth and Plan-etary Science Letters, v. 304, p. 520–526, doi: 10 .1016 /j .epsl .2011 .02 .028 .

Gallagher, K., Charvin, K., Nielsen, S., Sambridge, M., and Stephenson, J., 2009, Markov chain Monte Carlo (MCMC) sampling methods to determine optimal models, model resolution and model choice for earth science problems: Marine and Petroleum Geology, v. 26, p. 525–535, doi: 10 .1016 /j .marpetgeo .2009 .01 .003 .

Gehrels, G.E., 2000, Introduction to detrital zircon studies of Paleozoic and Triassic strata in western Nevada and northern California, in Soreghan, M.J., and Gehrels, G.E., eds., Paleozoic and Triassic Paleogeography and Tectonics of Western Nevada and Northern California: Geological Society of America Special Paper 347, p. 1–17, doi: 10 .1130 /0 -8137 -2347 -7 .1.

Gehrels, G., Yin, A., and Wang, X., 2003, Detrital-zircon geochronology of the northeastern Tibetan Plateau: Geological Society of America Bulletin, v. 115, p. 881–896, doi: 10 .1130 /0016 -7606 (2003)115 <0881: DGOTNT>2 .0 .CO;2 .

Gehrels, G., Kapp, P., DeCelles, P., Pullen, A., Blakey, R., Weislogel, A., Ding, L., Guynn, J., Martin, A., McQuarrie, N., and Yin, A., 2011, Detrital zircon geochronology of pre-Tertiary strata in the Tibetan-Himalayan orogen: Tectonics, v. 30, TC5016, doi: 10 .1029 /2011TC002868 .

Guo, Z., Ruddiman, W., Hao, Q., Wu, H., Qiao, Y., Zhu, R., Peng, S., Wei, J., Yuan, B., and Liu, T., 2002, Onset of Asian desertification by 22 Myr ago inferred from loess deposits in China: Nature, v. 416, p. 159–163, doi: 10 .1038 /416159a .

Gylesjö, S., and Arnold, E., 2006, Clay mineralogy of a red clay–loess sequence from Lingtai, the Chinese Loess Plateau: Global and Planetary Change, v. 51, p. 181–194, doi: 10 .1016 /j .gloplacha .2006 .03 .002 .

Han, W., Fang, X., Ye, C., Teng, X., and Zhang, T., 2014, Tibet forcing Quaternary stepwise enhancement of westerly jet and Central Asian aridification: Carbonate isotope records from deep drilling in the Qaidam salt playa, NE Tibet: Global and Planetary Change, v. 116, p. 68–75, doi: 10 .1016 /j .gloplacha .2014 .02 .006 .

Hoke, G., Schmitz, M., and Bowring, S., 2014, An ultrasonic method for isolating nonclay components from clay-rich material: Geochemistry Geophysics Geosystems, v. 15, p. 492–498, doi: 10 .1002 /2013GC005125 .

Jasra, A., Stephens, D., Gallagher, K., and Holmes, C., 2006, Bayesian mixture modelling in geochronology via Markov chain Monte Carlo: Mathematical Geology, v. 38, p. 269–300, doi: 10 .1007 /s11004 -005 -9019 -3 .

Jeong, G.Y., Hillier, S., and Kemp, R.A., 2011, Changes in mineralogy of loess–paleosol sections across the Chinese Loess Plateau: Quaternary Research, v. 75, p. 245–255, doi: 10 .1016 /j .yqres .2010 .09 .001 .

Kang, S., Lu, Y.C., and Wang, X.L., 2011, Closely-spaced recuperated OSL dating of the last interglacial paleosol in the southeastern margin of Chinese Loess Plateau: Quaternary Geochronology, v. 6, p. 480–490, doi: 10 .1016 /j .quageo .2011 .04 .004 .

Kang, S., Wang, X., and Lu, Y., 2013, Quartz OSL chronol-ogy and dust accumulation rate changes since the Last Glacial at Weinan on the southeastern Chinese Loess Plateau: Boreas, v. 42, p. 815–829.

Kapp, P., Pelletier, J., Rohrmann, A., Heermance, R., Rus-sell, J., and Ding, L., 2011, Wind erosion in the Qaidam Basin, Central Asia: Implications for tectonics, paleo-climate, and the source of the Loess Plateau: GSA Today, v. 21, p. 4–10, doi: 10 .1130 /GSATG99A .1 .

Kapp, P., Pullen, A., Pelletier, J.D., Russell, J., Goodman, P., and Cai, F., 2015, From dust to dust: Quaternary wind erosion of the Mu Us Desert and Loess Plateau, China: Geology, v. 43, p. 835–838, doi: 10 .1130 /G36724 .1 .

Kohfeld, K., and Harrison, S., 2003, Glacial-interglacial changes in dust deposition on the Chinese Loess Pla-teau: Quaternary Science Reviews, v. 22, p. 1859–1878, doi: 10 .1016 /S0277 -3791 (03)00166 -5 .

Lease, R., Burbank, D., Gehrels, G., Wang, Z., and Yuan, D., 2007, Signatures of mountain building: Detrital zir-con U/Pb ages from northeastern Tibet: Geology, v. 35, p. 239–242, doi: 10 .1130 /G23057A .1 .

Lease, R., Burbank, D., Hough, B., Wang, Z., and Yuan, D., 2012, Pulsed Miocene range growth in northeastern Tibet: Insights from Xunhua Basin magnetostratigra-phy and provenance: Geological Society of America Bulletin, v. 124, p. 657–677, doi: 10 .1130 /B30524 .1 .

Li, G., Pettke, T., and Chen, J., 2011, Increasing Nd isotopic ratio of Asian dust indicates progressive uplift of the north Tibetan Plateau since the middle Miocene: Geol-ogy, v. 39, p. 199–202, doi: 10 .1130 /G31734 .1 .

Licht, A., van Cappelle, M., Abels, H.A., Ladant, J., Trabucho-Alexandre, J., France-Lanord, C., Donnadieu, Y., Vandenberghe, J., Rigaudier, T., Lecuyer, C., Terry, D., Jr., Adriaens, R., Boura, A., Guo, Z., Aung Naing Soe, Dupont-Nivet, G., and Jaeger, J.-J., 2014, Asian monsoons in a late Eocene greenhouse world: Nature, v. 513, p. 501–506, doi: 10 .1038 /nature13704 .

Maher, B.A., and Thompson, R., 1991, Mineral magnetic record of the Chinese loess and paleosols: Geology, v. 19, p. 3–6, doi: 10 .1130 /0091 -7613 (1991)019 <0003: MMROTC>2 .3 .CO;2 .

Massey, F.J., Jr., 1951, The Kolmogorov-Smirnov test for goodness of fit: Journal of the American Statistical Asso ciation, v. 46, p. 68–78, doi: 10 .1080 /01621459 .1951 .10500769 .

Métivier, F., Gaudemer, Y., Tapponnier, P., and Meyer, B., 1998, Northeastward growth of the Tibet Plateau de-duced from balanced reconstruction of two deposi-tional areas: The Qaidam and Hexi Corridor basins, China: Tectonics, v. 17, p. 823–842, doi: 10 .1029 /98TC02764 .

Mroczek, P., 2013, Recycled loesses—A micromorphological approach to the determination of local source areas of Weichselian loess: Quaternary International, v. 296, p. 241–250, doi: 10 .1016 /j .quaint .2013 .02 .040 .

Nie, J., and Peng, W., 2014, Automated SEM-EDS heavy mineral analysis reveals no provenance shift between glacial loess and interglacial paleosol on the Chinese Loess Plateau: Aeolian Research, v. 13, p. 71–75, doi: 10 .1016 /j .aeolia .2014 .03 .005 .

Nie, J., Peng, W., Moller, A., Song, Y., Stockli, D.F., Stevens, T., Horton, B., Liu, S., Bird, A., Oalmann, J., Gong, H., and Fang, X., 2014, Provenance of the Upper Mio-cene–Pliocene red clay deposits of the Chinese Loess Plateau: Earth and Planetary Science Letters, v. 407, p. 35–47, doi: 10 .1016 /j .epsl .2014 .09 .026 .

Nie, J., Stevens, T., Rittner, M., Stockli, D., Garzanti, E., Limonta, M., et al., 2015, Loess plateau storage of northeastern Tibetan Plateau–derived Yellow River

Licht et al.


sediment: Nature Communications, v. 6, p. 8511, doi: 10 .1038 /ncomms9511 .

Peizhen, Z., Molnar, P., and Downs, W.R., 2001, Increased sedimentation rates and grain sizes 2–4 Myr ago due to the influence of climate change on erosion rates: Na-ture, v. 410, p. 891–897, doi: 10 .1038 /35073504 .

Porter, S., 2001, Chinese loess record of monsoon climate during the last glacial-interglacial cycle: Earth-Science Reviews, v. 54, p. 115–128, doi: 10 .1016 /S0012 -8252 (01)00043 -5 .

Porter, S., and An, Z., 1995, Correlation between climate events in the North Atlantic and China during the last glaciation: Nature, v. 375, p. 305–308, doi: 10 .1038 /375305a0 .

Pullen, A., Kapp, P., McCallister, A., Chang, H., Gehrels, G., Garzione, C., Heermance, R., and Ding, L., 2011, Quaidam Basin and northern Tibetan Plateau as dust sources for the Chinese Loess Plateau and paleocli-matic implications: Geology, v. 39, p. 1031–1034, doi: 10 .1130 /G32296 .1 .

Pullen, A., Ibanez-Mejia, M., Gehrels, G., Ibanez-Mejia, J., and Pecha, M., 2014, What happens when n = 1000? Creating large-n geochronological datasets with LA-ICP-MS for geologic investigations: Journal of Ana-lytical Atomic Spectrometry, v. 29, p. 971–980, doi: 10 .1039 /c4ja00024b .

Pye, K., 1995, The nature, origin and accumulation of loess: Quaternary Science Reviews, v. 14, p. 653–667, doi: 10 .1016 /0277 -3791 (95)00047 -X .

Qiang, X., An, Z., Song, Y., Chang, H., Sun, Y., Liu, W., Ao, H., Dong, J., Fu, C., Wu, F., Lu, F., Cai, Y., Zhou, W., Cao, J., Xu, X., and Ai, L., 2011, New eolian red clay sequence on the western Chinese Loess Plateau linked to onset of Asian desertification about 25 Ma ago: Sci-ence China, Earth Science, v. 54, p. 136–144.

Roe, G., 2009, On the interpretation of Chinese loess as a paleoclimate indicator: Quaternary Research, v. 71, p. 150–161, doi: 10 .1016 /j .yqres .2008 .09 .004 .

Rohrmann, A., Heermance, R., Kapp, P., and Cai, F., 2013, Wind as the primary driver of erosion in the Qaidam Basin, China: Earth and Planetary Science Letters, v. 374, p. 1–10, doi: 10 .1016 /j .epsl .2013 .03 .011 .

Song, Y., Fang, X., King, J.W., Li, J., Naoto, I., and An, Z., 2014, Magnetic parameter variations in the Chaona loess/paleosol sequences in the central Chinese Loess Plateau, and their significance for the middle Pleisto-cene climate transition: Quaternary Research, v. 81, p. 433–444, doi: 10 .1016 /j .yqres .2013 .10 .002 .

Stacey, J.S., and Kramers, J.D., 1975, Approximation of ter-restrial lead isotope evolution by a two-stage model: Earth and Planetary Science Letters, v. 26, p. 207–221, doi: 10 .1016 /0012 -821X (75)90088 -6 .

Stevens, T., and Lu, H., 2009, Optically stimulated lumi-nescence dating as a tool for calculating sedimentation rates in Chinese loess: Comparisons with grain-size re-cords: Sedimentology, v. 56, p. 911–934, doi: 10 .1111 /j .1365 -3091 .2008 .01004 .x .

Stevens, T., Palk, C., Carter, A., Lu, H., and Clift, P., 2010, Assessing the provenance of loess and desert sediments in northern China using U-Pb dating and morphology of detrital zircons: Geological Society of America Bul-letin, v. 122, p. 1331–1344, doi: 10 .1130 /B30102 .1 .

Stevens, T., Carter, A., Watson, T., Vermeesch, P., Ando, S., Bird, A., Lu, H., Garzanti, E., Cottam, M., and Sevast-janova, I., 2013, Genetic linkage between the Yellow River, the Mu Us Desert and the Chinese Loess Pla-teau: Quaternary Science Reviews, v. 78, p. 355–368, doi: 10 .1016 /j .quascirev .2012 .11 .032 .

Sun, D., Su, R., Bloemendal, J., and Lu, H., 2008, Grain-size and accumulation rate records from late Cenozoic aeolian sequences in northern China: Implications for variations in the East Asian winter monsoon and west-erly atmospheric circulation: Palaeogeography, Palaeo-climatology, Palaeoecology, v. 264, p. 39–53, doi: 10 .1016 /j .palaeo .2008 .03 .011 .

Sun, D., Zhang, Y., Han, F., Zhang, Y., Yi, Z., Li, Z., Wang, F., Wu, S., and Li, B., 2011, Magnetostratigraphy and palaeoenvironmental records for a late Cenozoic sedi-mentary sequence from Lanzhou, northeastern margin of the Tibetan Plateau: Global and Planetary Change, v. 76, p. 106–116, doi: 10 .1016 /j .gloplacha .2010 .12 .006 .

Sun, J., 2002, Provenance of loess material and formation of loess deposits on the Chinese Loess Plateau: Earth and Planetary Science Letters, v. 203, p. 845–859, doi: 10 .1016 /S0012 -821X (02)00921 -4 .

Sun, J., Zhang, M., and Liu, T., 2001, Spatial and tempo-ral characteristics of dust storms in China and its sur-rounding regions, 1960–1999: Relations to source area and climate: Journal of Geophysical Research, v. 106, p. 10,325–10,333, doi: 10 .1029 /2000JD900665 .

Sun, Y., Lu, H., and An, Z., 2006, Grain size of loess, palaeo-sol and red clay deposits on the Chinese Loess Plateau: Significance for understanding pedogenic alteration and palaeomonsoon evolution: Palaeogeography, Palaeo-climatology, Palaeoecology, v. 241, p. 129–138, doi: 10 .1016 /j .palaeo .2006 .06 .018 .

Sweeney, M.R., and Mason, J.A., 2013, Mechanisms of dust emission from Pleistocene loess deposits, Nebraska, USA: Journal of Geophysical Research–Earth Surface, v. 118, p. 1460–1471.

Toggweiler, J., and Russell, J., 2008, Ocean circulation in a warming climate: Nature, v. 451, p. 286–288, doi: 10 .1038 /nature06590 .

Vandenberghe, J., 2013, Grain size of fine-grained wind-blown sediment: A powerful proxy for process identi-fication: Earth-Science Reviews, v. 121, p. 18–30, doi: 10 .1016 /j .earscirev .2013 .03 .001 .

Vandenberghe, J., Renssen, H., van Huissteden, K., Nugteren, G., Konert, M., Lu, H., Dodonov, A., and Buylaert, J.-P., 2006, Penetration of Atlantic westerly winds into Central and East Asia: Quaternary Science Reviews, v. 25, p. 2380–2389, doi: 10 .1016 /j .quascirev .2006 .02 .017 .

Vermeesch, P., 2004, How many grains are needed for a provenance study?: Earth and Planetary Science Letters, v. 224, p. 441–451, doi: 10 .1016 /j .epsl .2004 .05 .037 .

Vermeesch, P., 2012, On the visualisation of detrital age dis-tributions: Chemical Geology, v. 312–313, p. 190–194, doi: 10 .1016 /j .chemgeo .2012 .04 .021 .

Vermeesch, P., 2013, Multi-sample comparison of detrital age distributions: Chemical Geology, v. 341, p. 140–146, doi: 10 .1016 /j .chemgeo .2013 .01 .010 .

Vermeesch, P., and Garzanti, E., 2015, Making geological sense of ‘Big Data’ in sedimentary provenance analy-sis: Chemical Geology, v. 409, p. 20–27, doi: 10 .1016 /j .chemgeo .2015 .05 .004 .

van Loon, A., 2006, Lost loesses: Earth-Science Reviews, v. 74, p. 309–316, doi: 10 .1016 /j .earscirev .2005 .10 .005 .

Wang, J., Wang, Y.J., Liu, Z.C., Li, J.Q., and Xi, P., 1999, Cenozoic environmental evolution of the Qaidam Ba-sin and its implications for the uplift of the Tibetan Pla-teau and the drying of Central Asia: Palaeogeography, Palaeoclimatology, Palaeoecology, v. 152, p. 37–47, doi: 10 .1016 /S0031 -0182 (99)00038 -3 .

Wang, L., Lü, H.Y., Wu, N.Q., Li, J., Pei, Y.P., Tong, G.B., and Peng, S.Z., 2006, Palynological evidence for late Miocene–Pliocene vegetation evolution recorded in the red clay sequence of the central Chinese Loess Pla-teau and implication for palaeoenvironmental change: Palaeogeography, Palaeoclimatology, Palaeoecology, v. 241, p. 118–128, doi: 10 .1016 /j .palaeo .2006 .06 .012 .

Wang, X., Dong, Z., Zhang, J., and Liu, L., 2004, Modern dust storms in China: An overview: Journal of Arid En-vironments, v. 58, p. 559–574, doi: 10 .1016 /j .jaridenv .2003 .11 .009 .

Weislogel, A., Graham, S., Chang, E., Wooden, J., and Geh-rels, G., 2010, Detrital zircon provenance from three turbidite depocenters of the Middle–Upper Triassic Songpan-Ganzi complex, central China: Record of col-lisional tectonics, erosional exhumation, and sediment production: Geological Society of America Bulletin, v. 122, p. 2041–2062, doi: 10 .1130 /B26606 .1 .

Wu, F., Fang, X., Ma, Y., Herrmann, M., Mosbrugger, V., An, Z., and Miao, Y., 2007, Plio–Quaternary stepwise drying of Asia: Evidence from a 3-Ma pollen record from the Chinese Loess Plateau: Earth and Planetary Science Letters, v. 257, p. 160–169, doi: 10 .1016 /j .epsl .2007 .02 .029 .

Xiao, G., Zong, K., Li, G., Hu, Z., Dupont-Nivet, G., Peng, S., and Zhang, K., 2012, Spatial and glacial-intergla-cial variations in provenance of the Chinese Loess Pla-teau: Geophysical Research Letters, v. 39, p. L20715, doi: 10 .1029 /2012GL053304 .

Yan, Y., Sun, Y., Chen, H., and Ma, L., 2014, Oxygen isotope signatures of quartz from major Asian dust sources: Implications for changes in the provenance of Chi-nese loess: Geochimica et Cosmochimica Acta, v. 139, p. 399–410, doi: 10 .1016 /j .gca .2014 .04 .043 .

Yin, A., Dang, Y.-Q., Zhang, M., Chen, X.-H., and McRivette, M.W., 2008, Cenozoic tectonic evolution of the Qaidam Basin and its surrounding regions (Part 3): Structural geology, sedimentation, and regional tec-tonic reconstruction: Geological Society of America Bulletin, v. 120, p. 847–876, doi: 10 .1130 /B26232 .1 .

Science editor: AAron cAvoSie ASSociAte editor: An Yin

MAnuScript received 30 June 2015 reviSed MAnuScript received 9 noveMber 2015 MAnuScript Accepted 15 deceMber 2015

Printed in the USA

Documents

Eolian cannibalism: Reworked loess and fluvial sediment as the … · 2018. 8. 7. · Reworked loess and fluvial sediment as the main sources of the Chinese Loess Plateau Geological